Hi, I'm Matthew Fontana.
I build the platforms that make AI coding tools actually work for engineering teams. I write here about what I'm learning along the way.
Currently Staff Engineer at Airbnb. Previously Spotify, UPS.
- Write here on agentic developer platforms, MCP, evals, and the agentic SDLC.
- Speak on podcasts, at conferences, and inside engineering orgs.
- Ship open source on GitHub (aitk, claude-code-toolkit, claude-tmux-manager, and other Claude Code tooling).
- Run TabletopAdventureCreator.com, a generative-AI SaaS in production since 2022.
- Help a handful of engineering teams a year build the platform layer underneath their AI tools.
From the blog
All posts
Guides & Fundamentals6 min readManaged Agents: The Production Layer Above Coding CLIs
Managed agents are the 2026 story. What they are, why they differ from CLIs like Claude Code and Codex, and a code-first POC built on Anthropic's platform.
Guides & Fundamentals9 min readBuild Your AI Harness on Marketplace Plugins, Not Vendors
Marketplace plugins ship five portable primitives — commands, skills, sub-agents, hooks, MCPs — that let you build business AI workflows across vendors.
Guides & Fundamentals8 min readWhere to Invest in AI Without Getting Vendor-Tied
Stop betting on AI vendors. Invest in three vendor-agnostic layers — tools, prompt workflows, and evaluation frameworks — that survive every model swap.
36 posts · 55,133 words · since October 2025. More always in the queue.